Stroke diagnosis and prognosis prediction method and system

1. A stroke diagnosis and prognosis prediction system in a stroke diagnosis and prognosis prediction method, the system comprising:

an image acquisition unit receiving a plurality of images comprising at least some of a human brain;

an image array unit arranging the plurality of images based on a standard brain image;

a lesion area detection and mapping unit detecting lesion areas in the plurality of images, respectively, and generating a single mapping image by mapping the plurality of images to the detected lesion areas;

a matching and correction unit matching the mapping image with the standard brain image by scaling the mapping image and performing image correction on the mapping image;

a three-dimensional image generation unit generating a three-dimensional lesion image by storing the mapping image in a three-dimensional data space; and

a stroke diagnosis unit diagnosing a stroke based on the three-dimensional lesion image,

wherein the stroke diagnosis unit includes:

a 3-D lesion image feature extraction unit extracting the features of the three-dimensional lesion image stored in the three-dimensional data space using a deep neural network;

a stroke cause classification unit diagnosing a stroke by classifying a stroke cause based on the extracted features of the three-dimensional lesion image;

a severity classification unit classifying the severity of the diagnosed stoke;

an intra-three-week risk prediction unit predicting a risk of the worsening possibility of a patient state within three weeks after the hospital stay of the patient as a percentage; and

a patient state prediction unit predicting a patient state after three months based on mRS classification regulations,

wherein the intra-three-week risk prediction unit classifies 90% or more of the rate of a risk as risk very high, classifies 70%˜90% of the rate of a risk as risk high, classifies 30%˜70% of the rate of a risk as risk normal, and classifies less than 30% of the rate of a risk as risk low.

2. The stroke diagnosis and prognosis prediction system of claim 1 , wherein the lesion area detection and mapping unit matches a mapped location of a lesion with the standard brain image based on matching location information of the plurality of images.

3. The stroke diagnosis and prognosis prediction system of claim 1 , wherein the three-dimensional image generation unit binarizes pixel information of the three-dimensional lesion image depending on whether a lesion is present with respect to the lesion image before the lesion image is generated, and adjusts a size of the lesion image.

4. The stroke diagnosis and prognosis prediction system of claim 1 , wherein the plurality of images comprises MRI images.

5. The stroke diagnosis and prognosis prediction system of claim 4 , wherein the MRI images comprise diffusion weighted imaging (DWI), a fluid attenuated inversion recovery (FLAIR) image, a gradient echo (GE) image, and a T2 weighted image (T2).

6. A stroke diagnosis and prognosis prediction method, comprising steps of:

obtaining a plurality of images comprising at least some of a human brain;

arranging the plurality of images based on a standard brain;

detecting lesion areas in the plurality of images, respectively, and generating a single mapping image by mapping the plurality of images to the detected lesion areas;

matching the mapping image with a standard brain image by scaling the mapping image and performing image correction on the mapping image;

generating a three-dimensional lesion image by storing the mapping image in a three-dimensional data space; and

extracting features of the three-dimensional lesion image using a deep neural network and diagnosing a stroke based on a deep neural network trained using the extracted three-dimensional lesion image,

wherein the extracting features of the three-dimensional lesion image step includes:

extracting the features of the three-dimensional lesion image stored in the three-dimensional data space using a deep neural network;

diagnosing a stroke by classifying a stroke cause based on the extracted features of the three-dimensional lesion image;

classifying the severity of the diagnosed stoke;

predicting a risk of the worsening possibility of a patient state within three weeks after the hospital stay of the patient as a percentage,

wherein the percentage classifies 90% or more of the rate of a risk as risk very high, classifies 70%˜90% of the rate of a risk as risk high, classifies 30%˜70% of the rate of a risk as risk normal, and classifies less than 30% of the rate of a risk as risk low; and

predicting a patient state after three months based on mRS classification regulations.

7. The stroke diagnosis and prognosis prediction method of claim 6 , wherein the step of generating the single mapping image comprises a step of matching a mapped location of the lesion with the standard brain image based on location information of the matching.

8. The stroke diagnosis and prognosis prediction method of claim 6 , further comprising steps of:

binarizing pixel information of the three-dimensional lesion image depending on whether a lesion is present with respect to the lesion image before the lesion image is generated, and

adjusting a size of the lesion image.